Help with data analysis

Forum for discussion on different signal processing algorithms
Locked
dyd1985
Posts: 23
Joined: 13 May 2010, 11:19

Help with data analysis

Post by dyd1985 » 27 May 2011, 13:00

Hi everybody.

I have a question... I already know that is almost impossible for you to give me accurate answers, considering you don't know much on my data acquisition system... anyway, I have attached the results of the offline analysis of recently acquired data. It was a long training session with several runs and trials. Task: 4 imagined movements (stimulus presentation on a Windows 7 (64bit) machine).

More details:
Condition 1 is left hand movement while 2 is right hand movement. No (digital) filters applied.
Channel order:
1: P3
2: T5
3: T3
4: C3
5: F3
6: Fp1
7: Fz
8: Cz
9: Pz
10: Fp2
11: F4
12: C4
13: T4
14: T6
15: P4

Can you please give me some hints on how to deal with these data? Why the spectra don't show any specific oscillations, while r^2 values show activity at some electrodes and some frequency.
I have carried out the same analysis using rest as first condition and as second condition either a left or a right imagined movement but the results were even worse.

Thanks,
Alessandro
Attachments
Features.jpg
Features.jpg (64.88 KiB) Viewed 3902 times
Spectra.jpg
Spectra.jpg (60.78 KiB) Viewed 3902 times

gschalk
Posts: 615
Joined: 28 Jan 2003, 12:37

Re: Help with data analysis

Post by gschalk » 30 May 2011, 15:08

Hi,

It seems as if the data are actually quite reasonable. R^2 of 0.2 with peaks at C3 and C4. I would do topographies in addition to the spectra, but you should see nice typical peaks over motor cortex.

There is considerable inter-subject variation in the topography and spectra of mu/beta rhythms, which is the reason why it's not so trivial to distinguish these signals from artifacts.

Gerv

dyd1985
Posts: 23
Joined: 13 May 2010, 11:19

Re: Help with data analysis

Post by dyd1985 » 31 May 2011, 09:51

Hello Dr. Schalk,
Thank you so much for your reply.
gschalk wrote:Hi,

It seems as if the data are actually quite reasonable. R^2 of 0.2 with peaks at C3 and C4. I would do topographies in addition to the spectra, but you should see nice typical peaks over motor cortex.

There is considerable inter-subject variation in the topography and spectra of mu/beta rhythms, which is the reason why it's not so trivial to distinguish these signals from artifacts.

Gerv
I'm sorry to bother you further, but I still don't understand the weird looking shape of the spectra graphs. r^2 values may seem reasonable as you noticed, but then why are the spectra, with respect to the two movements, so similar? I mean, by using only these plots it would be impossible to distinguish among different movements. In other words, I think I don't fully understand how r^2 values can be that high even if the respective spectra are not that different? I know this may seem a silly question, but I have also looked at your dissertation and yet I can't make up my mind.

Thanks
Alessandro

Locked

Who is online

Users browsing this forum: No registered users and 20 guests